FYI.

This story is over 5 years old.

Tech

Computers Can Predict Your Personality Based on Your Facebook Likes

And they know you better than your friends and family.

It's a thing we humans pride ourselves on, a thing that separates us from the robots: our varied and individual personalities. We're all just such precious unique snowflakes.

But we might be easier to figure out than we think. A new study published in PNAS reports that personality judgments made by a computer model can be more accurate than those made by a person's friends or family.

How did the model read people so accurately? Based on stuff they liked on Facebook.

Advertisement

"Previously we showed that you can use people's Facebook likes to predict what their personality is like," co-author David Stillwell, deputy director of Cambridge University's Psychometrics Centre, told Motherboard. "But where this study's different is that, whereas before we just showed you can make a prediction, this time we thought, how accurate is that really?"

The result: for the average person with 250 likes, the computer model was about as accurate as a real-life spouse. Family, friends, cohabitants, and work colleagues didn't perform as well.

table1.jpg

Judging personality sounds like a difficult ask; something you'd expect to require human social skills. But the pages you "like" reveal more than you think.

In the study, tens of thousands of volunteers filled in a personality test based on the established "Big Five" model, which looks at the broad traits of openness, conscientiousness, extraversion, agreeableness, and neuroticism (they're sometimes known as the OCEAN traits). Human judges filled out a longer, ten question version of the personality form, and could use whatever they knew about the subject to do so. The computer only had Facebook likes at its disposal, using a model trained on the personalities of a different sample of people and their corresponding likes. The model didn't use likes on statuses and pictures, but things such as books, movies, and bands that appear as "likes" on your Facebook profile (these used to be called "pages").

Advertisement

It might seem derivative—we don't like being put into boxes based on whether we like Star Wars or Britney Spears—but Stilwell said it is possible for the model to make predictions of personality based on Facebook likes. He gave me some examples for the trait of extraversion: Extroverted people like meeting new people (go figure), parties, beer pong, and making people laugh, but also less obvious things like the jeweller Tiffany & Co. Introverts, on the other hand, like mathematics, animé, Minecraft, Star Trek, and JRR Tolkien's The Silmarillion. "Many of them are rather stereotypical, but these are backed by data," he added.

On the other hand, if you want to be open to new experiences, you might want to get into Salvador Dali, meditation, and TED Talks.

I thought back to my own Facebook page and how I enthusiastically clicked a bunch of books and films when I first joined the site as a teenager but much fewer recently. Looking back, my likes include some pretentious foreign-language films, moody emo and folk bands, and a cat café. I knew where a computer model would put me on an extraversion scale—though I apparently do like "going out" too.

"For me it's not completely surprising, because if I went on a blind date with someone I'd never met before and they were telling me their favourite music and their favourite films, what sports they like playing, then I could make a judgment about what kind of person they are," said Stilwell. "But obviously what's different about our analysis is it's a computer doing this analysis automatically, without having any real understanding of, what is Big Brother?" It only knows what people who like Big Brother are generally like.

A computer model that can gauge your personality might personalize online interactions to your tastes, for example—something Stilwell compared to old-fashioned service where a shopkeeper would know you personally. Ideally, Stilwell said he would like to see more explanation as to why the online experience is currently tailored as it is. "I'd like to know that, the reason you're seeing Star Trek is because we think you're an introverted person," he said—and then offer you a chance to change that. You could also use a similar tool to help with tasks like matching people to careers that suit them.

The PNAS paper makes the inevitable reference to a Her-like scenario, and concludes, "The ability to accurately assess psychological traits and states, using digital footprints of behavior, occupies an important milestone on the path toward more social human-computer interactions."

Stilwell admitted that a real-life Samantha is a long way off, but this kind of tool nevertheless shows that computers can have some understanding of human psychology.

His work at the Psychometrics Centre pushes this further, and he suggested that we might not need to take a personality test like those used in this study at all in future. "Since so much of our behaviour is mediated by electronic devices—by cell phones, by the internet, by social networks—actually we could just measure this behaviour directly and we could get a much better way of understanding someone's personality, potentially, than even they can say themselves."